Contents lists available at ScienceDirect

Fuel

journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Investigation of ethanol oxidation over aluminum nanoparticle usingReaxFF molecular dynamics simulation

Yi Ran Zhanga, Adri C.T. van Duinb, Kai H. Luoa,c,⁎

a Center for Combustion Energy, Department of Energy and Power Engineering, Key Laboratory for Thermal Science and Power Engineering of Ministry of Education,Tsinghua University, Beijing 100084, ChinabDepartment of Mechanical and Nuclear Engineering, The Pennsylvania State University, University Park, PA 16802, USAc Department of Mechanical Engineering, University College London, Torrington Place, London WC1E 7JE, United Kingdom

A R T I C L E I N F O

Keywords:EthanolAluminum nanoparticle additivesReaxFF Molecular DynamicsNanofuelEnergetic fuel

A B S T R A C T

Aluminum nanoparticles are an effective and economical additive for producing energetic fuels. In the presentstudy, the state of the art ReaxFF molecular dynamics (MD) simulation has been used to uncover the detailedmechanisms of ethanol oxidation over aluminum nanoparticles with different oxidation states. The MD resultsreveal the dynamics process of ethanol oxidation reactions at nanoscales. The presence of aluminum nano-particles is found to reduce the initial temperature of ethanol oxidation to 324 K. It is also found that comparedto ethanol, oxygen molecules are more easily adsorbed on aluminum surfaces. Moreover, different oxidationstates of aluminum nanoparticles influence the initial ethanol reactions on the nanoparticles’ surfaces. OH-abstraction is more commonly observed on pure aluminum nanoparticles while H-abstraction prevails on alu-minum nanoparticles with oxide. The separated H atom from hydroxyl forms bonds with Al and O atom onaluminum nanoparticles surrounded by thin and thick oxide layers, respectively. Adsorptive dissociation ofethanol is hindered by the oxide layer surrounding the aluminum nanoparticle. Gas products like H2O and COresulting from ethanol oxidation on aluminum nanoparticles with the thick oxide layer are observed whilealmost all the C, H and O atoms in ethanol diffuse into the nanoparticles without or with the thin oxide layer. Forethanol dissociation, a higher temperature is required than adsorption. In addition, the rate of ethanol dis-sociation increases with rising reaction temperatures. The activation energy for ethanol adsorptive dissociationis found to be 4.58 kcal/mol on the aluminum nanoparticle with the thin oxide layer, which is consistent withresults from much more expensive DFT calculations.

1. Introduction

Metals have high energy density and are effective energetic ad-ditives in fuels and propellants [1–3]. Nano-sized metal particles showdecreased ignition delay and increased burning rates compared tomicro-sized particles [4–8]. Aluminum nanoparticles, in particular, canbe adopted as an energetic additive to enhance energy density andreduce the consumption of liquid fuels because of ready availability andlow cost. Recently, considerable efforts have been made to study theeffects of added aluminum nanoparticles on liquid fuel ignition andcombustion characteristics. Allen et al. [9] studied the ignition prop-erties of ethanol fuels with addition of 2%-wt aluminum nanoparticlesin a shock tube experiment. The results showed a significant reductionof the ignition delay time as compared to pure ethanol fuels. Tyagi et al.[10] found that aluminum and alumina additives to diesel fuels

enhance ignition probability significantly. Gan et al. [11] found sub-stantial enhancement in burning rates of ethanol due to the addition of80 nm aluminum particles. A plausible explanation for this effect is theenhanced thermal conductivity due to nanoparticle additives [12].However, this explanation cannot explain a contrary observation thatadding aluminum nanoparticles did not influence the burning rate ofJP-8 [13]. The above findings suggest that reactions between aluminumadditives and the base fuel should be taken into consideration, as alu-minum nanoparticles may function as catalysts. Aluminum nano-particles are not known as a good catalyst, but its catalytic effect mayimprove because of high specific surface areas.

Understanding of ethanol oxidation over aluminum nanoparticles oraluminum oxide nanoparticles is limited. The size, oxidation state andmorphology of the aluminum nanoparticles all affect the ethanol oxi-dation process. According to the DFT studies [14,15], the cleavage of

https://doi.org/10.1016/j.fuel.2018.06.119Received 30 April 2018; Received in revised form 20 June 2018; Accepted 28 June 2018

⁎ Corresponding author at: Center for Combustion Energy, Department of Energy and Power Engineering, Key Laboratory for Thermal Science and Power Engineering of Ministry ofEducation, Tsinghua University, Beijing 100084, China.

E-mail address: K.Luo@ucl.ac.uk (K.H. Luo).

Fuel 234 (2018) 94–100

0016-2361/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

T

the hydroxyl group of ethanol on Al2O3 (0 0 0 1) is the main reactionpathway for ethanol dissociation, and further decomposition of ethoxyhas a much higher energy barrier. Results from Alexander [16] showthat in the system of AlO and ethanol, the H-abstraction pathwayleading to formation of AlOH and C2H5O dominates. One problem withthe quantum mechanics (QM) method is that the computational cost isprohibitively high so that QM simulations are still limited to smallsystems (∼100 atoms) and small timescales (fs-ps).

All the above indicate the fundamental mechanisms of ethanolcombustion with aluminum nanoparticle additives are still unclear, dueto the complicated nature and a lack of effective research tools foratomic scale problems. An appealing alternative method is the reactivemolecular dynamics (MD) using ReaxFF force field, which can re-produce DFT reaction energies and barriers with a significantly reducedcomputational cost and describe chemical reactions in a much widerrange of scales than DFT [17,18]. The ReaxFF MD method has beenapplied to many systems, such as aluminum oxidation, hydrocarboncombustion and catalysis [19].

In this research, we use ReaxFF MD simulations to study the me-chanisms of ethanol oxidation with aluminum nanoparticle additiveswith different oxidation states on ethanol oxidation, and the funda-mental heterogeneous reaction mechanisms in general. The systems ofthe C2H5OH/O2 mixture reacting over aluminum nanoparticles withdifferent oxidation states are considered. Using ReaxFF MD, we are ableto look into both the physical and chemical processes of ethanol oxi-dation over aluminum nanoparticles at the atomic scales. The insightgained fills an important knowledge gap and helps the preparation ofaluminum nanoparticles for practical use. Reaction rates and activationenergies of ethanol dissociation over the aluminum nanoparticles aredetermined from simulation results quantitatively.

2. Methodology

ReaxFF-based reactive molecular dynamics is a powerful tool forsimulating heterogeneous reactions. ReaxFF is a general bond-orderbased force field, in which the connectivity of atoms is determined bybond orders calculated from interatomic distances that are updated atevery MD step. The bond order between a pair of atoms can be obtaineddirectly from interatomic distance. In calculating the bond orders,ReaxFF distinguished between contributions from sigma bonds, pi-bonds and double pi bonds. The force field is parameterized againstquantum mechanics based training sets and able to describe the bondformation and breaking which cannot be achieved by other non-re-active force fields [20]. Compared to QM calculations, ReaxFF MD al-lows simulations of reaction for longer physical time and larger sys-tems. The overall system energy expression is shown in Eq. (1).

= + + + + + + +E E E E E E E E Esystem bond over under 1p val tors vdWaals Coulomb

(1)

The total energy Esystem includes connectivity dependent terms suchas the bond energy (Ebond), overcoordination energy (Eover), under-coordination energy (Eunder), lone-pair energy (Elp), valence angle en-ergy (Eval) and torsion angle energy (Etors) [21]. The bond order (BO)between a pair of atoms can be obtained directly from interatomicdistance. In calculating the bond orders, ReaxFF distinguished betweencontributions from sigma bonds (BOij

σ), pi-bonds (BOijπ) and double pi

bonds (BOijππ). Here we use Ebond as an example to illustrate how the

bonded interaction is calculated:

= − − − −E D BO BO D BO· ·exp[p (1 (BO ) )] D ·bondσ

e ijσ σ π

ijπ

eππ

ijππ

bel ijP

ebe2 (2)

where the first part is the sigma bonds energy, the second part is the pi-bonds energy and the third part is the double pi bonds part. Further-more, ReaxFF describes non-bonded interactions such as van der Waalsand Coulomb interactions between all atoms, irrespective of con-nectivity [21]. Shielding terms have been used to avoid extremelyshort-range interactions. The ReaxFF also takes into account

polarization effects by using a geometry-dependent charge distributionderived from an electronegativity equalization method. Further in-formation about the ReaxFF formalism is available in previous studies[22]. The Al/C/H/O ReaxFF force field we used in our study has beendeveloped by Hong [24]. It has been successfully trained to high tem-perature reactions and applied to aluminum oxidation, aluminumcarbon coating and other Al containing systems [23–25].

ReaxFF MD simulations are conducted using canonical ensemble(NVT) [26,27] to prepare Al nanoparticles with oxide and to study theC2H5OH/O2 mixture reactions on Al nanoparticles. Simulations wereperformed on four different systems in a box size of equal measure-ments, which were 50× 50×50Å. Periodic boundary conditions wereimplemented in all three directions. The ReaxFF MD simulations arecarried out with the REAXC package in the LAMMPS platform [28].Visual Molecular Dynamics (VMD) [29] and OVITO [30] are used todisplay simulation results and the system configurations.

3. Results and discussion

3.1. Aluminum nanoparticle preparation

AnAl nanoparticle of a diameter 2.8 nm containing 856 atoms isfirstly prepared. The selection of this diameter is based on the objectiveof investigating aluminum melting temperature, oxidation process andcarbon coating process by ReaxFF molecular dynamics without ex-cessive computational cost. The prepared aluminum nanoparticle isthen placed in a simulation box that is measured 50×50×50Å.

Two NVT MD simulations with temperature of 298 K are performedto prepare aluminum nanoparticles with different oxide before in-vestigating ethanol oxidation reactions. 300 and 600 oxygen moleculeswere distributed randomly around the aluminum nanoparticle in thesimulation box. The initial system configurations are displayed inFig. 1. The velocities and positions are updated by the Velocity-Verletmethod. A time step of 0.2 fs for 2× 106 iterations (up to 0.4 ns) wasassigned because 0.2 fs describes the reactions of oxygen molecules onaluminum nanoparticles efficiently [25]. The changes in potential en-ergy and oxygen consumption over time are illustrated in Fig. 1. After0.3 ns, it is observed that the potential energy is stable and there is nomore oxygen consumption, which means the oxidation process isequilibrated and the aluminum nanoparticle is inert. A single aluminumnanoparticle with oxide can be produced by removing oxygen mole-cules in environmental gas. The oxide layer thicknesses are 0.76 nm and1.03 nm for the aluminum nanoparticles surrounded by 300 and 600oxygen molecules, respectively. The oxide layer thickness in our si-mulation is consistent with previous studies [23]. Here, we denote thepure aluminum nanoparticle AP, the aluminum nanoparticle with 300oxygen molecules APO300 and the aluminum nanoparticle with 600oxygen molecules APO600. The nanoparticles that were produced inthese simulations will be used in the later ethanol oxidation reactions.

3.2. Reaction of ethanol and oxygen mixture over aluminum nanoparticles

Allen et. al [9] have revealed that with addition of 2 wt% aluminumnanoparticles in ethanol fuel, a 32% reduction of ignition delay com-pared to pure ethanol fuel has been observed which could led to en-hancement in ignition probability. However, the experimental time-scales (∼10ms) are beyond MD simulation timescales due tocomputational cost consideration [31,32]. In our simulations, 20ethanol molecules and 60 oxygen molecules were added in the en-vironment, which made it a stoichiometric mixture. NVT moleculardynamics simulations are employed to study C2H5OH/O2 mixture re-actions over aluminum nanoparticles with different oxidation states asshown in Fig. 2. Ethanol and oxygen molecules were initially givenGaussian velocity distribution at 298 K. The nanoparticles with dif-ferent oxidation states were placed in the middle of three simulationboxes. The temperature is calculated by: = ×KE NKTdim /2

Y.R. Zhang et al. Fuel 234 (2018) 94–100

95

where KE is total kinetic energy of the system, dim is the di-mensionality of the simulation, N represents number of atoms in thegroup, k is the Boltzmann constant, and T is temperature. To comparethe influence of nanoparticles, the C2H5OH/O2 mixture without thepresence of any aluminum nanoparticle was simulated first. For theC2H5OH/O2 mixture system, the temperature was set at 298 K initiallyand was increased to 3000 K at a rate of 2 K/ps. The initial reaction wasobserved at 2610 K. The key reactions are illustrated in Fig. 3, in whichethanol molecules were decomposed into H2O and CH3CH radical first.Oxygen reacted with CH3OH radical to form CH3CHO (acetaldehyde)and O radicals. Then the O radical reacted with ethanol to produceCH3CH2O (ethoxy radical) and OH radical. The OH radicals also reactedwith ethanol to form CH3CH2O (ethoxy radical) and H2O.

For the aluminum nanoparticle systems, the temperature was in-itially set at 298 K and then increased to 500 K with the same heatingrate. Fig. 4 illustrates evolutions of the temperature, relative potentialenergy and number of ethanol molecules for the three systems withdifferent aluminum nanoparticles. In the pure aluminum nanoparticlesystem, the relative potential energy decrease and ethanol consumptionrates are faster than the other two systems, which indicates a fasterreaction rate in pure aluminum nanoparticle system. The ethanol oxi-dation reaction process for the C2H5OH/O2 mixture over the aluminumnanoparticle can be summarized in two steps: ethanol and oxygen

Fig. 1. (a) Initial configuration of a single aluminum nanoparticle surroundedby 300 oxygen molecules (b) Number of oxygen molecules consumed and po-tential energy decrease over time at 298 K.

Fig. 2. Initial configurations of ethanol oxidation reaction systems (a) C2H5OH/O2 mixture (b) AP surrounded by C2H5OH/O2 mixture (c) APO300 surrounded byC2H5OH/O2 mixture (d) APO600 surrounded by C2H5OH/O2 mixture (white= aluminum, red= oxygen green= carbon, yellow=hydrogen). (For interpretation ofthe references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3. Key reactions in ethanol oxidation at 3000 K.

Y.R. Zhang et al. Fuel 234 (2018) 94–100

96

molecules were first absorbed on the surface of the nanoparticle, andthen ethanol dissociated on the nanoparticle. At 500 K, the carbon-carbon bond in all the three systems did not break because the tem-perature is not high enough. For the pure aluminum nanoparticlesystem which will be referred to the “AP system”, the initial reactionwas observed at 324 K, when the OH in ethanol fell apart and formedbonds with the aluminum atom. In AP system, the OH-abstraction re-action dominates. By using bond-restraint method in ReaxFF, the acti-vation energy for OeH and CαeO bond cleavages energy barriers on Al(111) surface are 26.2 and 6.5 kcal/mol, which are in accordance toprevious DFT results that OH-abstraction reaction has the lowest energybarrier for atomic Al and ethanol system. The energy required foroxygen dissociation on aluminum surface is lower than that for ethanoldissociation. As a result, for the AP system at 500 K, oxygen moleculesdissociated while C2H5 molecules were still stuck on the aluminumsurface.

For the thin oxide aluminum nanoparticle system, which will bereferred to as the “APO300 system”, the initial ethanol dissociationreaction on the surface of the nanoparticle is different from the APsystem. At 330 K, the H atom in the hydroxy fell apart which resulted init bonding with the aluminum atom. For the thick oxide aluminumnanoparticle system, which will be referred to as the “APO600 system”,the first ethanol reaction was observed at 484 K, the H atom in thehydroxy fell apart but formed bonds with oxygen atoms on the surfaceof the nanoparticle instead of the aluminum atom. In ReaxFF, the OeHand CαeH bond clevages reaction energy barriers on Al2O3 (001)surface are 12.8 and 41.2 kcal/mol, which are close to previous DFTvalues of 18.7 and 57.6 kcal/mol. Aa a result, in APO300 and APO600systems, the OeH bond clevage dominates. The finding that H atom inthe hydroxyl fell apart confirms previous DFT studies [14]. In the DFTstudies of ethanol and AlO reaction, the H-abstraction activation energyleading to ethoxyl radical formation is lower than that required forhydroxyethyl radical formation. From ReaxFF results, the adsorptionenergies for oxygen and ethanol molecules on Al2O3 (001) surface are−13.2 and −7.8 kcal/mol. As a result, the oxygen molecules are

favorably adsorbed on aluminium nanoparticle surface. At 500 K, basedon the results of the simulations, we can conclude that with addition ofaluminium nanoparticle, a significant reduction of the initial ethanolreaction temperature was observed, which could potentially advancethe ethanol ignition. Additionally, for different aluminum nanoparticleswith different oxidation states, the initial ethanol oxidation reactionpathways on the surface of the aluminum nanoparticle are different asshown in Fig. 5.

With the same heating rate, the temperature is increased to 2000 Kfor the aluminum nanoparticle systems. The reactions observed at500 K were the OH or hydrogen separating from ethanol molecules andoxygen dissociation. At 2000 K, the initial ethanol oxidation reactionsare same with 500 K. For the AP system, the ethanol absorbed anddissociated on the surface of the nanoparticle completely. All the C andH atoms in the ethanol fell apart and diffused inside the Al nanoparticle.The adsorbed C2H5OH decomposed to C2H5 and OH, with H and Odiffusing inside the nanoparticle and the C2H5 sticking to Al. After that,the H atoms in C2H5 fell apart and formed bonds with aluminum atoms.Fig. 6 illustrates the evolution of the number of H and O atoms bothinside and on the surface of the nanoparticle. The H and O atoms in-creased first and then remained constant, which means the ethanolmolecules dissociated completely and the H and O atoms diffused insidenanoparticle. No product is found in the gas phase.

For the APO300 system, H in hydroxyl fell apart and formed bondwith aluminum atoms first, then diffused inside the nanoparticle. Afterthat, O and other H atoms in ethanol molecules fell apart and diffusedinside the nanoparticle. The C atoms were the last to diffuse inside thenanoparticle. At 2000 K, the temperature is high enough for H and O todiffuse and form OH radical. As a result, several H2 molecules and OHradicals were observed in the gas phase as illustrated in Fig. 7. Thetrend of evolution in the number of H and O atoms over time in theAPO300 system is similar to that in the AP system. The H and O atomsincreased and then remained almost constant, which means they diffuseinto the nanoparticle. The lower increase rate of H and O atoms

Fig. 4. Changes in relative potential energy, temperature and number ofethanol molecules as a function of time.

Fig. 5. Snapshots of C2H5OH adsorption and dissociation over (a) AP (b)APO300 (c) APO600.

Y.R. Zhang et al. Fuel 234 (2018) 94–100

97

compared to the AP system indicates a lower reaction rate.For the APO600 system at 2000 K, there are more O atoms on the

nanoparticle surface, so reactions and products were different from thepure and thin ANP systems. The H atom fell apart and formed bondswith oxygen atoms instead of aluminum atoms. After that, at 2000 K,the H and O atoms separated from ethanol were not able to diffuseinside the nanoparticle, resulting in the formation of more gas productslike H2O and CO as shown in Fig. 7. The number of O atoms increased

at first and then decreased, which indicates that oxygen atoms came outof the aluminum nanoparticle and formed O-containing species. Com-pared with the AP and APO300 system, more gas products were ob-served in the APO600 system. The number of H atoms fluctuated morethan the APO300 system. The difference confirms that the differentoxidation states of the aluminum nanoparticle influence the ethanolreactions on the aluminum nanoparticle surface.

From the simulation results, we can conclude that: first, aluminumnanoparticle additives in ethanol reduce the initial reaction tempera-ture, which consequently accelerates the ethanol ignition; and second,the different oxidation states of the aluminum nanoparticles influencethe ethanol reactions and products on the surface.

3.3. Acceleration of ethanol oxidation by phase transition of aluminum core

To investigate the effects of the nanoparticle on ethanol oxidation,we also conducted simulations in microcanonical ensemble (NVE) forAPO300 and APO600 systems with an initial temperature of 298 K. Asshown in Fig. 8, the potential energy decreased over time. Energy re-leased from ethanol oxidation caused the temperature increase. Thetemperature of the APO300 and APO600 systems increases up to2700 K and 2400 K, respectively. The temperature increases faster andearlier in the APO300system compared to the APO600 system, in-dicating that ethanol oxidation in APO300 system has a lower energybarrier. It is worth noting that the temperature increases in a very shorttime period (200–250 ps), which validate that the oxidation reaction isa highly exothermic process. The results demonstrate that the presenceof aluminum nanoparticle increases the heat release and acceleratesethanol oxidation, therefore decreasing the ignition delay from theatomic level. Fig. 8 shows that the nanoparticle oxide thickness in-creases at the same time period with temperature increase. The oxidethickness and temperature increase dramatically when the temperatureis beyond 700 K, which is the melting point of aluminum cluster [23].As a result, further diffusion occurs through the liquid aluminum andaccelerates the ethanol oxidation. The temperature range over the sharpincrease period is consistent with the previous finding that the micro-explosion regime of nanofuel with aluminum nanoparticle additivescombustion started at 700 K [11]. The growth of the oxide layer is bothinward and outward: inward because of the movement of oxygen atomstowards the interior of the nanoparticle, and outward because of themovement of aluminum atoms towards the oxide surface.

The increase of temperature and oxide thickness leads to disorder ofthe aluminum nanoparticle. Radical distribution function (RDF) is uti-lized to identify the nanoparticle structure information. RDF describeshow density varies as a function of distance from a reference aluminumatom, the first highest peak in Al-Al RDF represents the nearest Al-Albond length. The RDFs of APO300 at 500 K and 2000 K are shown in thesupplemental material (Fig. S1). The first peak height at 2000 K is lowerthan at 500 K, which indicates the transition from ordered to disorderedstructure. The aluminum core could change from solid state to liquidstate in the temperature increase range. The liquid state aluminum coremakes it easier for the H and O atoms to diffuse inside the aluminumnanoparticle, which could provide more active reaction sites on thesurface. Based on these findings, it can be concluded that the phasetransition of the aluminum core accelerates the ethanol oxidation,which indicates a possible transition from a diffusively-controlledcombustion process to a kinetically-controlled combustion process.These conclusions explain phenomena in ignition delays and gasifica-tion rates of nanofuel dependence on temperature.

3.4. Chemical kinetic parameters for C2H5OH dissociation

We use C2H5OH consumption rates to study the activation energy ofthe C2H5OH dissociation on aluminum nanoparticles. The systems weresimulated in NVT ensemble with temperatures varying in a range from400 K to 2000 K over the APO300. Fig. 9 represents C2H5OH

Fig. 6. The number of H and O atoms in the aluminum nanoparticle when thetemperature increases from 298 K to 2000 K. (a) AP system (b) APO300 system(c) APO600 system.

Fig. 7. Number of gas products from ethanol oxidation at 2000 K in APO300and APO600 systems.

Y.R. Zhang et al. Fuel 234 (2018) 94–100

98

concentration evolutions over time obtained from simulation at eachtemperature. At the initial reaction stages, the C2H5OH consumptionrates increases with increasing temperature.

According to chemical kinetics, the first-order rate constant k can becalculated from a linear fit to ln[C H OH] /[C H OH]2 5 t 2 5 0. The [C2H5OH]0and [C2H5OH]t represent the C2H5OH concentration initially and at anarbitrary moment. Arrhenius parameters for ethanol dissociation onAPO300 are calculated by the linear fit line from the Arrhenius plot inFig. 9. The activation energy for ethanol dissociation on APO300 is4.58 kcal/mol. Our ReaxFF simulation results are in accordance withquantum mechanics results, for the C2H5OH dissociation energy on AlOdiatomic is 1.96 kcal/mol [16]. Activation energy of ethanol decom-position over Al2O3 (001) surface by DFT calculations is 18.68 kcal/mol [14], which is also in agreement with our results.

4. Conclusion

ReaxFF molecular dynamics simulations have been conducted toinvestigate ethanol oxidation over aluminum nanoparticles with dif-ferent oxidation states at the atomic scale. In the C2H5OH/O2 mixtureand nanoparticle system, the presence of aluminum nanoparticles hasreduced the initial reaction temperature to 324 K. The reaction path-ways and products are different in aluminum nanoparticle systems withdifferent oxidation states. OH-abstraction is more commonly observedon the pure aluminum nanoparticle while H-abstraction prevails on thealuminum nanoparticle with oxide. The separated H atom from hy-droxyl forms bonds with Al and O atoms on aluminum nanoparticlewith thin and thick oxide layers, respectively. The dissociation of theadsorbed oxygen and ethanol molecules on the oxygen-coated surface isharder than on the pure aluminum nanoparticle surface. Gas productslike H2O and CO resulting from ethanol oxidation in the APO600system are observed. In contrast, almost all the C, H and O atoms inethanol diffuse inside pure and thin aluminum nanoparticle systems.Moreover, the ethanol dissociation requires a higher temperature thanadsorption. From the NVE simulations, we conclude that during theethanol oxidation over the aluminum nanoparticle, the solid state alu-minum core has changed to liquid state, which could make further Hand O atoms diffuse inside the nanoparticle more easily. It providesmore active reaction sites on the surface and therefore accelerates the

Fig. 8. Temperature, energy and oxide thickness evolutions in NVE simulations for (a) APO300 system (b) APO600 system.

Fig. 9. (a) C2H5OH concentration at various temperatures over APO300 (b)Arrhenius plot for the C2H5OH dissociation over APO300.

Y.R. Zhang et al. Fuel 234 (2018) 94–100

99

ethanol dissociation dramatically. The activation energy for C2H5OHadsorptive dissociation over the Al nanoparticle with a thin oxide layeris determined to be 4.58 kcal/mol, which is in agreement with resultsfrom much more expensive DFT calculations.

Acknowledgments

The authors thankfully acknowledge that simulations were partlyperformed on the Tsinghua High-Performance Parallel Computer andpartly on ARCHER funded under the EPSRC projects “UK Consortiumon Mesoscale Engineering Sciences (UKCOMES)” (Grant No. EP/L00030X/1) and “High Performance Computing Support for UnitedKingdom Consortium on Turbulent Reacting Flow (UKCTRF)” (GrantNo. EP/K024876/1).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in theonline version, at https://doi.org/10.1016/j.fuel.2018.06.119.

References

[1] Yetter RA, Risha GA, Son SF. Proc Combust Inst 2009;32:1819–38.[2] Granier JJ, Pantoya ML. Combust Flame 2004;138:373–83.[3] Dreizin EL. Prog Energy Combust Sci 2009;35:141–67.[4] Sundaram D, Yang V, Yetter RA. Prog Energy Combust Sci 2017;61:293–365.[5] Saxena V, Kumar N, Saxena VK. Renew Sust Energ Rev 2017;70:563–88.[6] Huang Y, Risha GA, Yang V, Yetter RA. Combust Flame 2009;156(1):5–13.[7] Venu H, Madhavan V. Fuel 2016;186:176–89.

[8] Jones M, Li CH, Afjeh A, Peterson GP. Nanoscale Res Lett 2011;6(1):246.[9] Allen C, Mittal G, Sung CJ, Toulson E, Lee T. Proc Combust Inst

2011;33(2):3367–74.[10] Tyagi H, et al. Nano Lett 2008;8:1410–6.[11] Gan Y, Qiao L. Combust Flame 2011;158(2):354–68.[12] Tanvir S, Qiao L. J Propul Power 2014;31(1):408–15.[13] Pfeil MA, Rosen SC, Yu YC, Anderson WE, Son SF. AIAA Paper (2010) Nashville,

TW; 2010.[14] Chiang HN, Nachimuthu S, Cheng YC, Damayanti NP, Jiang JC. Appl Surf Sci

2016;363:636–43.[15] Kostestkyy P, Yu J, Gorte RJ, Mpourmpakis G. Catal. Sci Technol

2014;4(11):3861–9.[16] Sharipov AS, Starik AM. J Phys Chem A 2015;119(17):3897–904.[17] Liu J, Wu J, Zhu J, et al. Fuel 2016;178:85–92.[18] Döntgen M, Przybylski-Freund M-D, Kröger LC, Kopp WA, Ismail AE, Leonhard K,

et al. J Chem Theory Comput 2015;11(6):2517–24.[19] Mao Q, Van Duin AC, Luo KH. Proc Combust Inst 2017;36(3):4339–46.[20] Wang H, Feng Y, Zhang X, et al. Fuel 2015;145:241–8.[21] Van Duin AC, Strachan A, Stewman S, Zhang Q, Xu X, Goddard WA. J Phys Chem A

2003;107(19):3803–11.[22] Chenoweth K, van Duin AC, Goddard WA. J Phys Chem A 2008;112:1040–53.[23] Hong S, van Duin AC. J Phys Chem C 2015;119(31):17876–86.[24] Hong S, van Duin AC. J Phys Chem C 2016;120(17):9464–74.[25] Hong S. Atomistic-scale investigation of the growth kinetics of aluminum oxide

layers on aluminum nanoparticles and germanium-based semiconductors using theReaxFF reactive force field [PhD thesis]. United States: The Pennsylvania StateUniversity, State College; 2016.

[26] Nosé S. J Chem Phys 1984;81(1):511–9.[27] Hoover WG. Phys Rev A 1985;31(3):1695.[28] Aktulga HM, Fogarty JC, Pandit SA, Grama AY. Parallel Comput

2012;38(4):245–59.[29] Humphrey W, Dalke A, Schulten K. J Mol Graphics 1996;14(1):33–8.[30] Stukowski A. Simul Mater Sci Eng 2009;18(1):015012.[31] Mao Q, Ren Y, Luo KH, Li S. J Phys Chem C 2015;119(51):28631–9.[32] Li G, Wang F, Wang J, et al. Fuel 2016;181:238–47.

Y.R. Zhang et al. Fuel 234 (2018) 94–100

100